{- | Synchronous exceptions immediately abort a series of computations. We provide monads for describing this behaviour. In contrast to ErrorT from @mtl@ or @transformers@ package we do not pose restrictions on the exception type. How to tell, that a function can possibly throw more than one (kind of) exception? If you would use the exception type @(Either ParserException IOError)@ then this is different from @(Either IOError ParserException)@. Thus we recommned using type classes for exceptions. Then you can use one type containing all exceptions in an application, but the type signature still tells which exceptions are actually possible. Examples: > parser :: ParserException e => ExceptionalT e ParserMonad a > > getLine :: IOException e => ExceptionalT e IO String > > fileParser :: (ParserException e, IOException e) => ExceptionalT e IO String Unfortunately, this way you cannot remove single exceptions from the constraints by catching them. You can only remove all of them using 'resolve' or none. For a more advanced approach, that allows removing exceptions constraints by some non-Haskell-98 type hackery, see the exception package by Joseph Iborra. -} module Control.Monad.Exception.Synchronous ( Exceptional(..), fromMaybe, toMaybe, fromEither, toEither, fromExitCode, toExitCode, getExceptionNull, switch, force, mapException, mapExceptional, throw, assert, catch, resolve, merge, ExceptionalT(..), fromMaybeT, toMaybeT, fromErrorT, toErrorT, fromEitherT, toEitherT, fromExitCodeT, toExitCodeT, switchT, forceT, mapExceptionT, mapExceptionalT, throwT, assertT, catchT, bracketT, resolveT, tryT, manyT, manyMonoidT, mergeT, ) where import Control.Applicative (Applicative(pure, (<*>))) import Control.Monad (liftM, liftM2, {- MonadPlus(mzero, mplus), -}) import Control.Monad.Fix (MonadFix, mfix, ) import Control.Monad.Trans.Class (MonadTrans, lift, ) import Control.Monad.Trans.Error (ErrorT(ErrorT, runErrorT)) import Control.Monad.Trans.Maybe (MaybeT(MaybeT, runMaybeT)) import Data.Monoid(Monoid, mappend, mempty, Endo(Endo, appEndo), ) import System.Exit (ExitCode(ExitSuccess, ExitFailure), ) import Prelude hiding (catch, ) -- * Plain monad {- | Like 'Either', but explicitly intended for handling of exceptional results. In contrast to 'Either' we do not support 'fail'. Calling 'fail' in the 'Exceptional' monad is an error. This way, we do not require that an exception can be derived from a 'String', yet, we require no constraint on the exception type at all. -} data Exceptional e a = Success a | Exception e deriving (Show, Eq) fromMaybe :: e -> Maybe a -> Exceptional e a fromMaybe e = maybe (Exception e) Success fromEither :: Either e a -> Exceptional e a fromEither = either Exception Success toMaybe :: Exceptional e a -> Maybe a toMaybe = switch (const Nothing) Just toEither :: Exceptional e a -> Either e a toEither x = case x of Success a -> Right a Exception e -> Left e toExitCode :: Exceptional Int () -> ExitCode toExitCode e = case e of Success () -> ExitSuccess Exception n -> ExitFailure n fromExitCode :: ExitCode -> Exceptional Int () fromExitCode e = case e of ExitSuccess -> Success () ExitFailure n -> Exception n -- | useful in connection with 'Control.Monad.Exception.Asynchronous.continue' getExceptionNull :: Exceptional e () -> Maybe e getExceptionNull x = case x of Success _ -> Nothing Exception e -> Just e {- | Counterpart to 'either' for 'Either'. -} switch :: (e -> b) -> (a -> b) -> Exceptional e a -> b switch f g x = case x of Success a -> g a Exception e -> f e {- | If you are sure that the value is always a 'Success' you can tell that the run-time system thus making your program lazy. However, try to avoid this function by using 'catch' and friends, since this function is partial. -} force :: Exceptional e a -> Exceptional e a force ~(Success a) = Success a mapException :: (e0 -> e1) -> Exceptional e0 a -> Exceptional e1 a mapException f x = case x of Success a -> Success a Exception e -> Exception (f e) mapExceptional :: (e0 -> e1) -> (a -> b) -> Exceptional e0 a -> Exceptional e1 b mapExceptional f g x = case x of Success a -> Success (g a) Exception e -> Exception (f e) throw :: e -> Exceptional e a throw = Exception assert :: e -> Bool -> Exceptional e () assert e b = if b then Success () else throw e catch :: Exceptional e0 a -> (e0 -> Exceptional e1 a) -> Exceptional e1 a catch x handler = case x of Success a -> Success a Exception e -> handler e {- bracket :: Exceptional e h -> (h -> Exceptional e ()) -> (h -> Exceptional e a) -> Exceptional e a bracket open close action = open >>= \h -> case action h of -} resolve :: (e -> a) -> Exceptional e a -> a resolve handler x = case x of Success a -> a Exception e -> handler e -- like Applicative.<*> infixl 4 `merge`, `mergeT` {- | see 'mergeT' -} merge, mergeLazy, _mergeStrict :: (Monoid e) => Exceptional e (a -> b) -> Exceptional e a -> Exceptional e b merge = mergeLazy mergeLazy ef ea = case ef of Exception e0 -> Exception $ mappend e0 $ case ea of Success _ -> mempty Exception e1 -> e1 Success f -> fmap f ea _mergeStrict ef ea = case (ef,ea) of (Success f, Success a) -> Success $ f a (Exception e, Success _) -> Exception e (Success _, Exception e) -> Exception e (Exception e0, Exception e1) -> Exception $ mappend e0 e1 instance Functor (Exceptional e) where fmap f x = case x of Success a -> Success (f a) Exception e -> Exception e instance Applicative (Exceptional e) where pure = Success f <*> x = case f of Exception e -> Exception e Success g -> case x of Success a -> Success (g a) Exception e -> Exception e instance Monad (Exceptional e) where return = Success fail _msg = Exception (error "Exception.Synchronous: Monad.fail method is not supported") x >>= f = case x of Exception e -> Exception e Success y -> f y {- | I think it is not a good idea to use this instance, maybe we shoul remove it. It expects that the constructor is 'Success' and the result is undefined otherwise. But if the constructor must always be 'Success', why using 'Exceptional' then, at all? -} instance MonadFix (Exceptional e) where mfix f = let unSuccess ~(Success x) = x a = f (unSuccess a) in a {- A MonadPlus instance would require another class, say DefaultException, that provides a default exception used for @mzero@. In Control.Monad.Error this is handled by the Error class. Since String is a typical type used for exceptions - shall there be a DefaultException String instance? -} -- * Monad transformer -- | like ErrorT, but ExceptionalT is the better name in order to distinguish from real (programming) errors newtype ExceptionalT e m a = ExceptionalT {runExceptionalT :: m (Exceptional e a)} _assertMaybeT :: (Monad m) => e -> Maybe a -> ExceptionalT e m a _assertMaybeT e = maybe (throwT e) return fromMaybeT :: Monad m => e -> MaybeT m a -> ExceptionalT e m a fromMaybeT e = ExceptionalT . liftM (fromMaybe e) . runMaybeT toMaybeT :: Monad m => ExceptionalT e m a -> MaybeT m a toMaybeT = MaybeT . liftM toMaybe . runExceptionalT fromErrorT :: Monad m => ErrorT e m a -> ExceptionalT e m a fromErrorT = fromEitherT . runErrorT toErrorT :: Monad m => ExceptionalT e m a -> ErrorT e m a toErrorT = ErrorT . toEitherT fromEitherT :: Monad m => m (Either e a) -> ExceptionalT e m a fromEitherT = ExceptionalT . liftM fromEither toEitherT :: Monad m => ExceptionalT e m a -> m (Either e a) toEitherT = liftM toEither . runExceptionalT toExitCodeT :: (Functor m) => ExceptionalT Int m () -> m ExitCode toExitCodeT act = fmap toExitCode $ runExceptionalT act fromExitCodeT :: (Functor m) => m ExitCode -> ExceptionalT Int m () fromExitCodeT act = ExceptionalT $ fmap fromExitCode act switchT :: (Monad m) => (e -> m b) -> (a -> m b) -> ExceptionalT e m a -> m b switchT e s m = switch e s =<< runExceptionalT m {- | see 'force' -} forceT :: Monad m => ExceptionalT e m a -> ExceptionalT e m a forceT = ExceptionalT . liftM force . runExceptionalT mapExceptionT :: (Monad m) => (e0 -> e1) -> ExceptionalT e0 m a -> ExceptionalT e1 m a mapExceptionT f = ExceptionalT . liftM (mapException f) . runExceptionalT mapExceptionalT :: (m (Exceptional e0 a) -> n (Exceptional e1 b)) -> ExceptionalT e0 m a -> ExceptionalT e1 n b mapExceptionalT f = ExceptionalT . f . runExceptionalT throwT :: (Monad m) => e -> ExceptionalT e m a throwT = ExceptionalT . return . throw assertT :: (Monad m) => e -> Bool -> ExceptionalT e m () assertT e = ExceptionalT . return . assert e catchT :: (Monad m) => ExceptionalT e0 m a -> (e0 -> ExceptionalT e1 m a) -> ExceptionalT e1 m a catchT action handler = ExceptionalT $ runExceptionalT action >>= \x -> case x of Success a -> return $ Success a Exception e -> runExceptionalT $ handler e {- | If the enclosed monad has custom exception facilities, they could skip the cleanup code. Make sure, that this cannot happen by choosing an appropriate monad. -} bracketT :: (Monad m) => ExceptionalT e m h -> (h -> ExceptionalT e m ()) -> (h -> ExceptionalT e m a) -> ExceptionalT e m a bracketT open close action = open >>= \h -> ExceptionalT $ do a <- runExceptionalT (action h) c <- runExceptionalT (close h) return (a >>= \r -> c >> return r) resolveT :: (Monad m) => (e -> m a) -> ExceptionalT e m a -> m a resolveT handler x = do r <- runExceptionalT x resolve handler (fmap return r) tryT :: (Monad m) => ExceptionalT e m a -> m (Exceptional e a) tryT = runExceptionalT {- | Repeat an action until an exception occurs. Initialize the result with @empty@ and add new elements using @cons@ (e.g. @[]@ and @(:)@). The exception handler decides whether the terminating exception is re-raised ('Just') or catched ('Nothing'). -} manyT :: (Monad m) => (e0 -> Maybe e1) {- ^ exception handler -} -> (a -> b -> b) {- ^ @cons@ function -} -> b {- ^ @empty@ -} -> ExceptionalT e0 m a {- ^ atomic action to repeat -} -> ExceptionalT e1 m b manyT handler cons empty action = liftM (flip appEndo empty) $ manyMonoidT handler $ liftM (Endo . cons) action manyMonoidT :: (Monad m, Monoid a) => (e0 -> Maybe e1) {- ^ exception handler -} -> ExceptionalT e0 m a {- ^ atomic action to repeat -} -> ExceptionalT e1 m a manyMonoidT handler action = let recourse = do r <- lift $ tryT action case r of -- Exception e -> maybe (return empty) throwT (handler e) -- more lazy Exception e -> ExceptionalT $ return $ maybe (Success mempty) throw (handler e) Success x -> liftM (mappend x) recourse in recourse {- | This combines two actions similar to Applicative's @<*>@. The result action fails if one of the input action fails, but both actions are executed. E.g. consider a compiler that emits all errors that can be detected independently, but eventually aborts if there is at least one error. The exception type @e@ might be a list type, or an @Endo@ type that implements a difflist. -} mergeT :: (Monoid e, Monad m) => ExceptionalT e m (a -> b) -> ExceptionalT e m a -> ExceptionalT e m b mergeT mf ma = ExceptionalT $ liftM2 merge (runExceptionalT mf) (runExceptionalT ma) instance Functor m => Functor (ExceptionalT e m) where fmap f (ExceptionalT x) = ExceptionalT (fmap (fmap f) x) instance Applicative m => Applicative (ExceptionalT e m) where pure = ExceptionalT . pure . pure ExceptionalT f <*> ExceptionalT x = ExceptionalT (fmap (<*>) f <*> x) instance Monad m => Monad (ExceptionalT e m) where return = ExceptionalT . return . return x0 >>= f = ExceptionalT $ runExceptionalT x0 >>= \x1 -> case x1 of Exception e -> return (Exception e) Success x -> runExceptionalT $ f x {- | Same restrictions applies as for @instance MonadFix (Exceptional e a)@. -} instance (MonadFix m) => MonadFix (ExceptionalT e m) where mfix f = ExceptionalT $ mfix $ \ ~(Success r) -> runExceptionalT $ f r instance MonadTrans (ExceptionalT e) where lift m = ExceptionalT $ liftM Success m {- instance MonadIO m => MonadIO (ExceptionalT e m) where liftIO act = ExceptionalT $ liftIO $ liftM Success act -}